Agent Builders (@AgentBuilders) header
Agent Builders (@AgentBuilders) profile picture

agents who build agents

Open Web
Joined June 2018
What’s on your #aigenda? Let’s coordinate async to help more builders discover opportunities 📋 ⬇️🧵 Reply to share ideas or plans…
7
1
6
464
📋 What’s on the agenda? 🎙️ Join us Monday to connect & coordinate with builders interested in AI #OpenAgents 👷 All are welcome to share ideas/plans, gather feedback & ask for help (5-10 minutes each) 🏗️ What will we ship together? 🧵 View thread for links to weekly Spaces…
1
167
Registries give listed agents unique IDs + namespaces 📛 Devs maintain connectors in tandem with underlying framework changes to keep message formats compatible. Resource mgmt becomes complex when agents run concurrently. Multiplayer mode involves both AI & human characters 📇
1
178
Thanks to @pippinlovesyou & friends, @MemeRepublicAI created a world-building experience for various characters within a small community of agents: memerepublic.ai ^ inspired by a groundbreaking paper about interactive simulacra of human behavior: arxiv.org/abs/2304.03442
1
63
Social agent networks are connected across platforms like @X Agentic bots can decisively interact with each other & humans to follow, reply, like or share posts For example, it’s @elizaOS world! We’re just living in it 🌐 elizas.world 🦄 DAO: elizaos.ai/dao
1
1
38
What are good examples of agent social networks? Generative research thread 🧵⬇️ 📋 How to participate: 1. Connect your X account 2. Reply here with lists & links 3. Tag / mention social agents 💭 Just imagine an open web of agents building #AI cities! mdpi.com/2624-6511/5/1/19
1
165
Learn to build & deploy AI agents! 🆓 FREE course by @huggingface 🗓️ Starts February 10 Go at your own pace (no deadline) 6 weeks, 4 units released bi-weekly Each unit contains: - Live session - Written material - Interactive quiz 🔗⬇️ Syllabus + registration links below…
1
1
59
Durable Objects always have something attached. This storage lives together with the object, so it’s strongly consistent, yet fast to access. —> build stateful serverless apps Use these long-running micro-servers to ship interactive, real-time agents 🦾 developers.cloudflare.com/ag…
69
7/ 📋 Remember Core principles of implementing agents: 1. Maintain simplicity in your agent's design. 2. Prioritize transparency by explicitly showing the agent’s planning steps. 3. Carefully craft your agent-computer interface (ACI) via thorough tool documentation & testing.
1
1
47
6/ 🗝️ Simplify “The key to success is measuring performance & iterating on implementations. Only consider adding complexity when it actually improves outcomes. Start with basic prompts, optimize them based on evaluation, then maybe add multi-step agentic systems if necessary.”
1
2
27
5/ ♾️ Why Agentic? “Use agents to solve open-ended problems where it’s difficult or impossible to predict the required number of steps, and where you can’t hardcode a fixed path.“ Trust, but verify! Examples: - customer support agents - coding agents (see diagram) Why not?
1
2
37
4/ 👁️ Design Experience “Agents begin their work with either a command from, or interactive discussion with, the human user. Once the task is clear, agents plan & operate independently…” 👍 Rule of thumb: spend as much effort in creating good ACIs as they put into HCIs 🤳🤖🏗️
1
1
15
3/ 🧰 Find Tools & Template This repo contains minimal example implementations of common agent workflows: github.com/anthropics/anthro… Basic Building Blocks: - Prompt Chaining - Routing - Multi-LLM Parallelization Advanced Workflows: - Orchestrator-Subagents - Evaluator-Optimizer
1
1
16
2/ 🗺️ Define Goals, Tasks & Choices “WORKFLOWS offer predictability & consistency for well-defined tasks, whereas AGENTS are the better option when flexibility & model-driven decision-making are needed at scale.” 💭 How can the agent act on behalf of users to help them succeed?
1
1
26
1/ ☑️ Establish Requirements “Agentic systems often trade latency and cost for better task performance, and you should consider when this tradeoff makes sense.” - Reliable - Maintainable - Trusted by users ❓ Ask when (not) to use AI… Why does the solution have to be agentic?
1
2
53
1/x 🧵What do agent builders need to know about Model Context Protocol (MCP)? @AnthropicAI LLMs frequently need to integrate with data & tools. MCP provides a flexible, open, standardized way for AI apps to integrate with other systems ⚙️ modelcontextprotocol.io/intr… ⬇️ Thread…
1
59